An in silico HBV model predicts viral response to the oral non-steroidal carboxylic acid FXR agonist EYP001a.
Monteiro Sousa Claudio (1), Jacob Evgueni (1) , Martin Bastien (1) , Laheux Samuel (1), Sampson Diane (2), Elise Roy (2), Jacky Vonderscher (2), Pietro Scalfaro (2), Patrice André (3), Boissel Jean-Pierre (1)
(1) Novadiscovery, Lyon, France, (2) ENYO pharma, Lyon, France, (3) CIRI - INSERM U1111 – CNRS UMR5308 - Université Lyon 1- ENS de Lyon
Objectives: In line with guidance to FDA to expand the use of trial simulations to support drug development [1], a mathematical model was built to simulate the effect of EYP001a treatment on hepatitis B virus (HBV) replication. EYP001a is a synthetic non-bile acid Farnesoid X receptor (FXR) agonist currently under clinical development for chronic HBV infection and NASH. FXR regulates bile acid metabolism and is a target for liver disease therapies. We aimed at exploring the EYP001a effect on hepatocyte virus and viral markers production.
Methods: The mechanistic model was based on curated knowledge extracted from white and grey scientific literature via the community-driven knowledge management platform (https://githealth.io). The complete model (105 ODEs, 287 parameters) integrated bile acids physiology, cholesterol metabolism, HBV replication and compound mode of action (the latter from EYP001a non-clinical data). The computational model was written and implemented through Novadiscovery’s proprietary simulation framework and its various tools (SimWork). The SimWork virtual population and exploration tools were used for calibrating the model: virtual patients were randomly generated from ranges of descriptor (representing the n model parameters) values and were selected on the basis of a score translating physiological and biological constraints that the model should comply with; this results in a n-dimension space domain where the parameter values meet the constraints. SimWork was also used for the generation of 10,000 HBV infected virtual patients. An independent team, blinded to available clinical EYP001a data, simulated the effect of single and multiple oral doses of EYP001a in healthy or chronically infected HBV virtual subjects. EYP001a effects on FXR response markers 7α-Hydroxy-4-cholesten-3-one (C4) and Fibroblast growth factor 19 (FGF19) were explored. Model performance was tuned with data observed in healthy subjects and simulated results were validated with data from both in vitro HepaRG experiments and in vivo HBV infected subjects. The effect on HBV replication of several combinations of EYP001a dosing regimens were explored. Additionally, different associated daily dietary intakes of cholesterol schemes were tested. HBV DNA and the surface antigens of the hepatitis B virus (HBsAg) output curves were generated for 10,000 virtual HBV subjects treated during 100 days with EYP001a.
Results: The model successfully predicted EYP001a plasma concentration-time profiles (Cmax, Tmax and AUC) at the different tested doses and regimens. The model reproduced accurately the dynamics of blood viral particles, C4 and FGF19 and their changes after single and multiple EYP001a administrations and predicted HBsAg levels. Comparison of in silico HBV DNA and HBsAg outputs indicated that prolonged, but not short lasting FXR agonism with EYP001a inhibited viral replication. Various combinations of dosing regimens with associated cholesterol dietary intakes were tested and it was established that 200mg EYP001a BID was an appropriate efficacious regimen.
Conclusions: The in silico model predicted well EYP001a plasma concentrations and pharmacodynamic response in a virtual HBV infected population. The strong predictability of our simulation approach using in silico modeling could be used to determine an a priori better dosing regimen in chronic HBV patients. This in silico model will be used to explore other FXR agonist treatment strategies and to identify best responders in the population to be tested in phase II HBV trials.
References:
[1] Pras Pathmanathan, Richard A. Gray, Vicente J. Romero, Tina M. Morrison, 2017, 'Applicability Analysis of Validation Evidence for Biomedical Computational Models', Journal of Verification, Validation and Uncertainty Quantification, vol. 2, no. 2, pp. 021005-undefined (doi:10.1115/1.4037671)
